Costs
Google Analytics 4 attribution models in 2027: what US teams must know
GA4 ships two attribution models and a reporting identity setting that decides channel credit. Lookback windows, counting and US-to-Canada flows change the totals.
What to take away
- GA4 gives you two reporting attribution models, data-driven and last click, plus a reporting identity setting that decides which users merge before credit is assigned.
- Lookback window and conversion counting move channel totals without moving a single dollar of spend.
- Blended identity pulls in Google Signals and modeled data, so consent and transfer rules apply to what you report.
- US teams with Canadian traffic need a documented cross-border transfer assessment, not only a cookie banner.
The two attribution models GA4 actually offers
Google Analytics 4 kept a narrower model set than Universal Analytics did. The Attribution settings page under Admin offers data-driven attribution and last click, and nothing else. The older first-click, linear, position-based and time-decay models no longer appear in GA4 reporting.
Data-driven attribution is a counterfactual model. It estimates how much a conversion depended on each touchpoint by comparing paths that included that touchpoint with paths that did not. Credit is then spread across paid and organic channels. Google Analytics background explains how the property and its measurement setup fit together.
Last click gives everything to the final touchpoint that met the conversion criteria.
Lookback, counting and identity settings
Three settings sit beside the model choice. Each one moves the channel totals you report, and each one is easy to leave on a default nobody chose.
GA4 attribution settings
Setting
- Attribution model
- Data-driven, last click
- Lookback window
- 30 or 90 days
- Conversion counting
- Once per event, once per session
- Reporting identity
- Blended, observed, device-based
Options
- Attribution model
- How credit is split
- Lookback window
- How far back a conversion looks
- Conversion counting
- Total conversion volume
- Reporting identity
- Which users are stitched together
What it changes
- Attribution model
- Lookback window
- Conversion counting
- Reporting identity
| Setting | Options in GA4 | What it changes |
|---|---|---|
| Reporting attribution model | Data-driven, last click | How credit is split |
| Lookback window | 30 or 90 days for acquisition, 30, 60 or 90 days for other conversions | How far back a conversion looks |
| Conversion counting | Once per event, once per session | Total conversion volume |
| Reporting identity | Blended, observed, device-based | Which users are stitched together |
Reporting identity changes channel credit
Blended identity stitches users using User ID first, then Google Signals, then device ID. Observed identity uses User ID then device ID. Device-based identity uses the device ID alone, so cross-device journeys split apart.
Switching from blended to device-based can move credit between paid social and branded search on the same day. It also changes user counts, because the reported unit changes with the setting. A page on why audience counts diverge shows how the unit of measurement, not the audience, decides the number.
Example: one signup, three different reports
A B2B signup sees a paid search ad on day 1, a webinar on day 20, and a demo request on day 45. With a 30 day lookback and last click, paid search earns nothing.
One signup, three reports
- Day 1Paid search ad
- Day 20Webinar
- Day 45Demo request
Set the lookback to 90 days and switch to data-driven attribution, and the webinar and paid search split the credit. Same user, same conversion, different answer. The sales process stage where each touch happens is what drives the gap.
Where data-driven attribution stops helping
Data-driven attribution needs volume. Conversions with few recorded paths do not give the model enough comparisons, so the property falls back to last click for that event. A low-traffic property can sit on last click for months without anyone noticing.
Sparse data is not the only limit. The model only sees signals inside the GA4 property and the Google products linked to it. Renewals, call center orders and long enterprise cycles stay invisible unless you import them from a CRM.
Cross-border flows for US teams with Canadian traffic
A US-owned GA4 property collecting Canadian visitor data is transferring personal information across a border. Canada's federal law, PIPEDA, applies to commercial activity with a real and substantial connection to Canada, as the PIPEDA overview sets out.
Quebec's Law 25 adds a transfer assessment duty. Before sending personal information outside Quebec, an organization must assess whether it will receive adequate protection in the destination. Data residency planning for analytics platforms follows the same reasoning, and data sovereignty concepts cover the wider argument.
Course coverage of these duties is uneven across providers. A review of Quebec Law 25 course checks lists the consent, retention and breach reporting topics to confirm before you pay for Canadian analytics training.
A settings review checklist
- Confirm the reporting attribution model matches how your team actually allocates budget.
- Set the lookback window to fit the sales cycle rather than accepting the default.
- Record which reporting identity is live, and the date it last changed.
- Check Google Signals consent status before relying on blended numbers.
- Document the transfer assessment for any Canadian traffic in the property.
- Compare GA4 channel credit against a holdout test or a mix model at least once a year.
Where GA4 sits beside other methods
Attribution in GA4 answers which touchpoints appeared before a conversion. It does not answer what would have happened without them. A set of attribution examples across nine funnels shows how those two questions differ across ecommerce, B2B, retail, subscriptions and lead generation.
Most US teams run GA4 attribution beside incrementality tests, marketing mix models and CRM reporting. Deciding which evidence wins an argument is a strategy choice. A guide to marketing analytics strategy choices for 2027 connects business questions to defined evidence, reliable data and documented learning.
